The application of sub-titles is an important element in video stream production. Sub-titles enable a user to have a better understanding of the content of the video, and particularly of the speech uttered, when the latter is hardly, if at all, understandable. This is particularly useful when a program is transmitted in a language not known to the user or when the auditory perception of the language is disturbed by poor transmission, poorly articulated speech, or background noise. Sub-titles also enable the hard of hearing to achieve an understanding of the program.
The application of sub-titles may be performed either offline or in real time (so-called “live” sub-titles in the parlance of the art). The application of sub-titles offline is performed on a support prepared in advance. This type of sub-titling is found, for example, in DVDs, Blu-ray discs, or transmissions broadcast offline. This offline application does not present any particular time constraint. Thus, it is possible in this case to manually insert sub-titles into the video stream, an operator verifying that the sub-titles are perfectly synchronized with the video, while presenting a visual aspect perceived to be pleasant by the user. Although subjective, this notion of pleasant perception of the sub-titles can be reduced to objective elements, for example, obtaining sub-titles which are displayed at a moderate speed, or which retain a fixed position with respect to the screen. It is also possible to use audio analysis schemes which may turn out to be expensive with respect to calculation time for purposes of best synchronizing the sub-titles with the audio/video.
On the other hand, it is not possible to use these techniques in the case of live sub-titles, as in this case the video content produced is transmitted immediately. The production of live sub-titles is generally performed by an operator producing the sub-titles live and dispatching them in the stream. This operation produces an inevitable lag between the moment at which the video is transmitted and the moment at which the corresponding sub-title is produced. This lag is still more significant when the task of the operator induces a translation from one language to the other for the production of the sub-titles.
The approach generally used to process live sub-titles is to transmit each word of sub-titles as soon as it is available in order not to add any further lag. However, this approach has drawbacks, e.g., in addition to an inevitable lag, the words arrive one after another, sometimes in a very fast manner. The live sub-titles are thus often perceived by users as unpleasant and of poor quality. The speed of appearance of the words forming sub-titles has been listed as one of the main causes of poor perception of live sub-titles by users by Ofcom (Office of Communications), The quality of live subtitling, pp. 29-30. According to this study, users seem to prefer sub-titles presented in blocks.
Critical speeds for the display of sub-titles are notably studied in DTV4A11, 2010. D2.5: Final Report on Pilot Services (2 of 2), UK Evaluation Report (Roehampton), pp. 6 to 9. This document shows notably the evaluation of quality of the sub-titles as a function of the number of words per minute (also abbreviated to “wpm” in the technical field concerned). The study shows that sub-titles comprising 180 words per minute are assessed as being of “good quality” by a much more significant percentage of users than sub-titles comprising 220 words per minute.